Running Speed Estimation Using Shoe-Worn Inertial Sensors: Direct Integration, Linear, and Personalized Model

Area of Science:

• Biomechanics
• Sports Science
• Wearable Technology

Background:

• Global Navigation Satellite System (GNSS) devices are commonly used for estimating running speed but suffer from signal issues and high power consumption.
• Inertial sensors offer a potential alternative for accurate and efficient speed estimation in running analysis.

Purpose of the Study:

• To propose and compare three novel methods for estimating overground running speed using foot-worn inertial sensors.
• To evaluate the performance of these methods against a reference GNSS system.

Main Methods:

• Three approaches were developed: direct strapdown integration of foot acceleration, a feature-based linear model, and a personalized online model using recursive least squares.
• Performance was assessed using data from 33 individuals running at various speeds in a real-world setting.
• Feature sets included both literature-based and automatically selected features.

Main Results:

• Direct integration showed an accuracy of 0.08 ± 0.1 m/s.
• The best feature-based linear model achieved 0.00 ± 0.11 m/s accuracy.
• The personalized model demonstrated superior performance with 0.00 ± 0.01 m/s accuracy and 0.09 ± 0.06 m/s precision.

Conclusions:

• Direct estimation of foot velocity is biased by overground velocity and slope.
• General linear models are limited by inter-subject gait variations.
• Personalized models effectively minimize inter-subject bias for accurate running speed estimation.